xref: /llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.sample.g16.ll (revision e7900e695e7dfb36be8651d914a31f42a5d6c634) 
1; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -stop-after=legalizer -o - %s | FileCheck -check-prefix=GFX10 %s
3; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -stop-after=legalizer -o - %s | FileCheck -check-prefix=GFX11 %s
4; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -stop-after=legalizer -o - %s | FileCheck -check-prefix=GFX12 %s
5
6define amdgpu_ps <4 x float> @sample_d_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, float %s) {
7  ; GFX10-LABEL: name: sample_d_1d
8  ; GFX10: bb.1.main_body:
9  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2
10  ; GFX10-NEXT: {{  $}}
11  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
12  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
13  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
14  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
15  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
16  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
17  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
18  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
19  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
20  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
21  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
22  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
23  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
24  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
25  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
26  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
27  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
28  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
29  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
30  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
31  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
32  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
33  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
34  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
35  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
36  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
37  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
38  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
39  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
40  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
41  ;
42  ; GFX11-LABEL: name: sample_d_1d
43  ; GFX11: bb.1.main_body:
44  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2
45  ; GFX11-NEXT: {{  $}}
46  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
47  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
48  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
49  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
50  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
51  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
52  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
53  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
54  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
55  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
56  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
57  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
58  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
59  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
60  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
61  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
62  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
63  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
64  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
65  ; GFX11-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
66  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
67  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
68  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
69  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
70  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
71  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
72  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
73  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
74  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
75  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
76  ;
77  ; GFX12-LABEL: name: sample_d_1d
78  ; GFX12: bb.1.main_body:
79  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2
80  ; GFX12-NEXT: {{  $}}
81  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
82  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
83  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
84  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
85  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
86  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
87  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
88  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
89  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
90  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
91  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
92  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
93  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
94  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
95  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
96  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
97  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
98  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
99  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
100  ; GFX12-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
101  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
102  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
103  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
104  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
105  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
106  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
107  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
108  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
109  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
110  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
111main_body:
112  %v = call <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
113  ret <4 x float> %v
114}
115
116define amdgpu_ps <4 x float> @sample_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
117  ; GFX10-LABEL: name: sample_d_2d
118  ; GFX10: bb.1.main_body:
119  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
120  ; GFX10-NEXT: {{  $}}
121  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
122  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
123  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
124  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
125  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
126  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
127  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
128  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
129  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
130  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
131  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
132  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
133  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
134  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
135  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
136  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
137  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
138  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
139  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
140  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
141  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
142  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
143  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
144  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
145  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
146  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
147  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
148  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
149  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
150  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
151  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
152  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
153  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
154  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
155  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
156  ;
157  ; GFX11-LABEL: name: sample_d_2d
158  ; GFX11: bb.1.main_body:
159  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
160  ; GFX11-NEXT: {{  $}}
161  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
162  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
163  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
164  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
165  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
166  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
167  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
168  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
169  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
170  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
171  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
172  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
173  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
174  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
175  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
176  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
177  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
178  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
179  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
180  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
181  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
182  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
183  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
184  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
185  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
186  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
187  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
188  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
189  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
190  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
191  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
192  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
193  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
194  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
195  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
196  ;
197  ; GFX12-LABEL: name: sample_d_2d
198  ; GFX12: bb.1.main_body:
199  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
200  ; GFX12-NEXT: {{  $}}
201  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
202  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
203  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
204  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
205  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
206  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
207  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
208  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
209  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
210  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
211  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
212  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
213  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
214  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
215  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
216  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
217  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
218  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
219  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
220  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
221  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
222  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
223  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
224  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
225  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
226  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
227  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
228  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
229  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
230  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
231  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
232  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
233  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
234  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
235  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
236main_body:
237  %v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
238  ret <4 x float> %v
239}
240
241define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r) {
242  ; GFX10-LABEL: name: sample_d_3d
243  ; GFX10: bb.1.main_body:
244  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
245  ; GFX10-NEXT: {{  $}}
246  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
247  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
248  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
249  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
250  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
251  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
252  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
253  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
254  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
255  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
256  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
257  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
258  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
259  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
260  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
261  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
262  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
263  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
264  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
265  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
266  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
267  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
268  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
269  ; GFX10-NEXT:   [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
270  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
271  ; GFX10-NEXT:   [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
272  ; GFX10-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
273  ; GFX10-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
274  ; GFX10-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
275  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
276  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
277  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
278  ; GFX10-NEXT:   [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[TRUNC4]](s16)
279  ; GFX10-NEXT:   [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[DEF]](s16)
280  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
281  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
282  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY20]](s32)
283  ; GFX10-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<14 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
284  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.3d), 15, [[CONCAT_VECTORS]](<14 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
285  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
286  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
287  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
288  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
289  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
290  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
291  ;
292  ; GFX11-LABEL: name: sample_d_3d
293  ; GFX11: bb.1.main_body:
294  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
295  ; GFX11-NEXT: {{  $}}
296  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
297  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
298  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
299  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
300  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
301  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
302  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
303  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
304  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
305  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
306  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
307  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
308  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
309  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
310  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
311  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
312  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
313  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
314  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
315  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
316  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
317  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
318  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
319  ; GFX11-NEXT:   [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
320  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
321  ; GFX11-NEXT:   [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
322  ; GFX11-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
323  ; GFX11-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
324  ; GFX11-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
325  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
326  ; GFX11-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
327  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
328  ; GFX11-NEXT:   [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[TRUNC4]](s16)
329  ; GFX11-NEXT:   [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[DEF]](s16)
330  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
331  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
332  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY20]](s32)
333  ; GFX11-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
334  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.3d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[CONCAT_VECTORS]](<6 x s16>), $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
335  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
336  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
337  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
338  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
339  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
340  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
341  ;
342  ; GFX12-LABEL: name: sample_d_3d
343  ; GFX12: bb.1.main_body:
344  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
345  ; GFX12-NEXT: {{  $}}
346  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
347  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
348  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
349  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
350  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
351  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
352  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
353  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
354  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
355  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
356  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
357  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
358  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
359  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
360  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
361  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
362  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
363  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
364  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
365  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
366  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
367  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
368  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
369  ; GFX12-NEXT:   [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
370  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
371  ; GFX12-NEXT:   [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
372  ; GFX12-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
373  ; GFX12-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
374  ; GFX12-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
375  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
376  ; GFX12-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
377  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
378  ; GFX12-NEXT:   [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[TRUNC4]](s16)
379  ; GFX12-NEXT:   [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[DEF]](s16)
380  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
381  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
382  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY20]](s32)
383  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR5]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
384  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.3d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[CONCAT_VECTORS]](<8 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
385  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
386  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
387  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
388  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
389  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
390  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
391main_body:
392  %v = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
393  ret <4 x float> %v
394}
395
396define amdgpu_ps <4 x float> @sample_c_d_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s) {
397  ; GFX10-LABEL: name: sample_c_d_1d
398  ; GFX10: bb.1.main_body:
399  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
400  ; GFX10-NEXT: {{  $}}
401  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
402  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
403  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
404  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
405  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
406  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
407  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
408  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
409  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
410  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
411  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
412  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
413  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
414  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
415  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
416  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
417  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
418  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
419  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
420  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
421  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
422  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
423  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
424  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
425  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
426  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
427  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
428  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
429  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
430  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
431  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
432  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
433  ;
434  ; GFX11-LABEL: name: sample_c_d_1d
435  ; GFX11: bb.1.main_body:
436  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
437  ; GFX11-NEXT: {{  $}}
438  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
439  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
440  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
441  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
442  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
443  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
444  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
445  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
446  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
447  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
448  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
449  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
450  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
451  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
452  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
453  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
454  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
455  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
456  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
457  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
458  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
459  ; GFX11-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
460  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
461  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
462  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
463  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
464  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
465  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
466  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
467  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
468  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
469  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
470  ;
471  ; GFX12-LABEL: name: sample_c_d_1d
472  ; GFX12: bb.1.main_body:
473  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
474  ; GFX12-NEXT: {{  $}}
475  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
476  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
477  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
478  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
479  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
480  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
481  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
482  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
483  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
484  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
485  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
486  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
487  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
488  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
489  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
490  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
491  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
492  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
493  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
494  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
495  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
496  ; GFX12-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
497  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
498  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
499  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
500  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
501  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
502  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
503  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
504  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
505  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
506  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
507main_body:
508  %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
509  ret <4 x float> %v
510}
511
512define amdgpu_ps <4 x float> @sample_c_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
513  ; GFX10-LABEL: name: sample_c_d_2d
514  ; GFX10: bb.1.main_body:
515  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
516  ; GFX10-NEXT: {{  $}}
517  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
518  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
519  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
520  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
521  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
522  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
523  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
524  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
525  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
526  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
527  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
528  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
529  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
530  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
531  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
532  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
533  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
534  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
535  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
536  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
537  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
538  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
539  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
540  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
541  ; GFX10-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
542  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
543  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
544  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
545  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
546  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
547  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
548  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
549  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
550  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
551  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
552  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
553  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
554  ;
555  ; GFX11-LABEL: name: sample_c_d_2d
556  ; GFX11: bb.1.main_body:
557  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
558  ; GFX11-NEXT: {{  $}}
559  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
560  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
561  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
562  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
563  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
564  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
565  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
566  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
567  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
568  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
569  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
570  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
571  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
572  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
573  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
574  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
575  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
576  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
577  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
578  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
579  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
580  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
581  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
582  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
583  ; GFX11-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
584  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
585  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
586  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
587  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
588  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
589  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
590  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
591  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
592  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
593  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
594  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
595  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
596  ;
597  ; GFX12-LABEL: name: sample_c_d_2d
598  ; GFX12: bb.1.main_body:
599  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
600  ; GFX12-NEXT: {{  $}}
601  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
602  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
603  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
604  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
605  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
606  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
607  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
608  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
609  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
610  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
611  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
612  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
613  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
614  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
615  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
616  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
617  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
618  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
619  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
620  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
621  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
622  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
623  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
624  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
625  ; GFX12-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
626  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
627  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
628  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
629  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
630  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
631  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
632  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
633  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
634  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
635  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
636  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
637  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
638  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
639main_body:
640  %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
641  ret <4 x float> %v
642}
643
644define amdgpu_ps <4 x float> @sample_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, float %s, float %clamp) {
645  ; GFX10-LABEL: name: sample_d_cl_1d
646  ; GFX10: bb.1.main_body:
647  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
648  ; GFX10-NEXT: {{  $}}
649  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
650  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
651  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
652  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
653  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
654  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
655  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
656  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
657  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
658  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
659  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
660  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
661  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
662  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
663  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
664  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
665  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
666  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
667  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
668  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
669  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
670  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
671  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
672  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
673  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
674  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
675  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
676  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
677  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
678  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
679  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
680  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
681  ;
682  ; GFX11-LABEL: name: sample_d_cl_1d
683  ; GFX11: bb.1.main_body:
684  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
685  ; GFX11-NEXT: {{  $}}
686  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
687  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
688  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
689  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
690  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
691  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
692  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
693  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
694  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
695  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
696  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
697  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
698  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
699  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
700  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
701  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
702  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
703  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
704  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
705  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
706  ; GFX11-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
707  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
708  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
709  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
710  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
711  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
712  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
713  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
714  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
715  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
716  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
717  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
718  ;
719  ; GFX12-LABEL: name: sample_d_cl_1d
720  ; GFX12: bb.1.main_body:
721  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
722  ; GFX12-NEXT: {{  $}}
723  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
724  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
725  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
726  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
727  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
728  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
729  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
730  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
731  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
732  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
733  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
734  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
735  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
736  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
737  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
738  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
739  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
740  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
741  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
742  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
743  ; GFX12-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
744  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
745  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
746  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
747  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
748  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
749  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
750  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
751  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
752  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
753  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
754  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
755main_body:
756  %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
757  ret <4 x float> %v
758}
759
760define amdgpu_ps <4 x float> @sample_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
761  ; GFX10-LABEL: name: sample_d_cl_2d
762  ; GFX10: bb.1.main_body:
763  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
764  ; GFX10-NEXT: {{  $}}
765  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
766  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
767  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
768  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
769  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
770  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
771  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
772  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
773  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
774  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
775  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
776  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
777  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
778  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
779  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
780  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
781  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
782  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
783  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
784  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
785  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
786  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
787  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
788  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
789  ; GFX10-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
790  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
791  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
792  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
793  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
794  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
795  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
796  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
797  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
798  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
799  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
800  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
801  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
802  ;
803  ; GFX11-LABEL: name: sample_d_cl_2d
804  ; GFX11: bb.1.main_body:
805  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
806  ; GFX11-NEXT: {{  $}}
807  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
808  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
809  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
810  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
811  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
812  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
813  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
814  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
815  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
816  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
817  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
818  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
819  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
820  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
821  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
822  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
823  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
824  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
825  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
826  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
827  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
828  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
829  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
830  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
831  ; GFX11-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
832  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
833  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
834  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
835  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
836  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
837  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
838  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
839  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
840  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
841  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
842  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
843  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
844  ;
845  ; GFX12-LABEL: name: sample_d_cl_2d
846  ; GFX12: bb.1.main_body:
847  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
848  ; GFX12-NEXT: {{  $}}
849  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
850  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
851  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
852  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
853  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
854  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
855  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
856  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
857  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
858  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
859  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
860  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
861  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
862  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
863  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
864  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
865  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
866  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
867  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
868  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
869  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
870  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
871  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
872  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
873  ; GFX12-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
874  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
875  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
876  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
877  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
878  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
879  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
880  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.d.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
881  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
882  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
883  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
884  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
885  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
886  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
887main_body:
888  %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
889  ret <4 x float> %v
890}
891
892define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
893  ; GFX10-LABEL: name: sample_c_d_cl_1d
894  ; GFX10: bb.1.main_body:
895  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
896  ; GFX10-NEXT: {{  $}}
897  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
898  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
899  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
900  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
901  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
902  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
903  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
904  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
905  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
906  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
907  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
908  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
909  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
910  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
911  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
912  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
913  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
914  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
915  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
916  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
917  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
918  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
919  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
920  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
921  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
922  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
923  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
924  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.cl.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
925  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
926  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
927  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
928  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
929  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
930  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
931  ;
932  ; GFX11-LABEL: name: sample_c_d_cl_1d
933  ; GFX11: bb.1.main_body:
934  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
935  ; GFX11-NEXT: {{  $}}
936  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
937  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
938  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
939  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
940  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
941  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
942  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
943  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
944  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
945  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
946  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
947  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
948  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
949  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
950  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
951  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
952  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
953  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
954  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
955  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
956  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
957  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
958  ; GFX11-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
959  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
960  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
961  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
962  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
963  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.cl.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
964  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
965  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
966  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
967  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
968  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
969  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
970  ;
971  ; GFX12-LABEL: name: sample_c_d_cl_1d
972  ; GFX12: bb.1.main_body:
973  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
974  ; GFX12-NEXT: {{  $}}
975  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
976  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
977  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
978  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
979  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
980  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
981  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
982  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
983  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
984  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
985  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
986  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
987  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
988  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
989  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
990  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
991  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
992  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
993  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
994  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
995  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
996  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
997  ; GFX12-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
998  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
999  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1000  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1001  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1002  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
1003  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.cl.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[CONCAT_VECTORS]](<4 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1004  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1005  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
1006  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1007  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1008  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1009  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1010main_body:
1011  %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
1012  ret <4 x float> %v
1013}
1014
1015define amdgpu_ps <4 x float> @sample_c_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
1016  ; GFX10-LABEL: name: sample_c_d_cl_2d
1017  ; GFX10: bb.1.main_body:
1018  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
1019  ; GFX10-NEXT: {{  $}}
1020  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1021  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1022  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1023  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1024  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1025  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1026  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1027  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1028  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1029  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1030  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1031  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1032  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1033  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1034  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1035  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1036  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1037  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1038  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1039  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1040  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1041  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1042  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
1043  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1044  ; GFX10-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
1045  ; GFX10-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
1046  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1047  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1048  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1049  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1050  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
1051  ; GFX10-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
1052  ; GFX10-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
1053  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.cl.2d), 15, [[CONCAT_VECTORS]](<12 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1054  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1055  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
1056  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1057  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1058  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1059  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1060  ;
1061  ; GFX11-LABEL: name: sample_c_d_cl_2d
1062  ; GFX11: bb.1.main_body:
1063  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
1064  ; GFX11-NEXT: {{  $}}
1065  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1066  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1067  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1068  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1069  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1070  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1071  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1072  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1073  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1074  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1075  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1076  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1077  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1078  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1079  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1080  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1081  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1082  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1083  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1084  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1085  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1086  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1087  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
1088  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1089  ; GFX11-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
1090  ; GFX11-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
1091  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1092  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1093  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1094  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1095  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
1096  ; GFX11-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
1097  ; GFX11-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
1098  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.cl.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1099  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1100  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
1101  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1102  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1103  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1104  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1105  ;
1106  ; GFX12-LABEL: name: sample_c_d_cl_2d
1107  ; GFX12: bb.1.main_body:
1108  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
1109  ; GFX12-NEXT: {{  $}}
1110  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1111  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1112  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1113  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1114  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1115  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1116  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1117  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1118  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1119  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1120  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1121  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1122  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1123  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1124  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1125  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1126  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1127  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1128  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1129  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1130  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1131  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1132  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
1133  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1134  ; GFX12-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
1135  ; GFX12-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
1136  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1137  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1138  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1139  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1140  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
1141  ; GFX12-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
1142  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
1143  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.cl.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[CONCAT_VECTORS]](<6 x s16>), $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1144  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1145  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
1146  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1147  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1148  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1149  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1150main_body:
1151  %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
1152  ret <4 x float> %v
1153}
1154
1155define amdgpu_ps <4 x float> @sample_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, float %s) {
1156  ; GFX10-LABEL: name: sample_cd_1d
1157  ; GFX10: bb.1.main_body:
1158  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2
1159  ; GFX10-NEXT: {{  $}}
1160  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1161  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1162  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1163  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1164  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1165  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1166  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1167  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1168  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1169  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1170  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1171  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1172  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1173  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1174  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1175  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1176  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1177  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1178  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1179  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1180  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1181  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1182  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
1183  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1184  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1185  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
1186  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1187  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1188  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1189  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1190  ;
1191  ; GFX11-LABEL: name: sample_cd_1d
1192  ; GFX11: bb.1.main_body:
1193  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2
1194  ; GFX11-NEXT: {{  $}}
1195  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1196  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1197  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1198  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1199  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1200  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1201  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1202  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1203  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1204  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1205  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1206  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1207  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1208  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1209  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1210  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1211  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1212  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1213  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1214  ; GFX11-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1215  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1216  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1217  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
1218  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1219  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1220  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
1221  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1222  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1223  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1224  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1225  ;
1226  ; GFX12-LABEL: name: sample_cd_1d
1227  ; GFX12: bb.1.main_body:
1228  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2
1229  ; GFX12-NEXT: {{  $}}
1230  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1231  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1232  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1233  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1234  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1235  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1236  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1237  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1238  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1239  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1240  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1241  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1242  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1243  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1244  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1245  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1246  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1247  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1248  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1249  ; GFX12-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1250  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1251  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1252  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
1253  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1254  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1255  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
1256  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1257  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1258  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1259  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1260main_body:
1261  %v = call <4 x float> @llvm.amdgcn.image.sample.cd.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
1262  ret <4 x float> %v
1263}
1264
1265define amdgpu_ps <4 x float> @sample_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
1266  ; GFX10-LABEL: name: sample_cd_2d
1267  ; GFX10: bb.1.main_body:
1268  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
1269  ; GFX10-NEXT: {{  $}}
1270  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1271  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1272  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1273  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1274  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1275  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1276  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1277  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1278  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1279  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1280  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1281  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1282  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1283  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1284  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1285  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1286  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1287  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1288  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1289  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1290  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1291  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1292  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1293  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1294  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1295  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1296  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1297  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1298  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1299  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1300  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
1301  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1302  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1303  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1304  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1305  ;
1306  ; GFX11-LABEL: name: sample_cd_2d
1307  ; GFX11: bb.1.main_body:
1308  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
1309  ; GFX11-NEXT: {{  $}}
1310  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1311  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1312  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1313  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1314  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1315  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1316  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1317  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1318  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1319  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1320  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1321  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1322  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1323  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1324  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1325  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1326  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1327  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1328  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1329  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1330  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1331  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1332  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1333  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1334  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1335  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1336  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1337  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1338  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1339  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1340  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
1341  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1342  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1343  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1344  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1345  ;
1346  ; GFX12-LABEL: name: sample_cd_2d
1347  ; GFX12: bb.1.main_body:
1348  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
1349  ; GFX12-NEXT: {{  $}}
1350  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1351  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1352  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1353  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1354  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1355  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1356  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1357  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1358  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1359  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1360  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1361  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1362  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1363  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1364  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1365  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1366  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1367  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1368  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1369  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1370  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1371  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1372  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1373  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1374  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1375  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1376  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1377  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1378  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1379  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1380  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
1381  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1382  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1383  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1384  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1385main_body:
1386  %v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
1387  ret <4 x float> %v
1388}
1389
1390define amdgpu_ps <4 x float> @sample_c_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s) {
1391  ; GFX10-LABEL: name: sample_c_cd_1d
1392  ; GFX10: bb.1.main_body:
1393  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
1394  ; GFX10-NEXT: {{  $}}
1395  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1396  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1397  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1398  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1399  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1400  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1401  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1402  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1403  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1404  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1405  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1406  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1407  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1408  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1409  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1410  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1411  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1412  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1413  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1414  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1415  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1416  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1417  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1418  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1419  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1420  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1421  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1422  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
1423  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1424  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1425  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1426  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1427  ;
1428  ; GFX11-LABEL: name: sample_c_cd_1d
1429  ; GFX11: bb.1.main_body:
1430  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
1431  ; GFX11-NEXT: {{  $}}
1432  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1433  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1434  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1435  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1436  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1437  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1438  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1439  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1440  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1441  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1442  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1443  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1444  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1445  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1446  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1447  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1448  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1449  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1450  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1451  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1452  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1453  ; GFX11-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1454  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1455  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1456  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1457  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1458  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1459  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
1460  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1461  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1462  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1463  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1464  ;
1465  ; GFX12-LABEL: name: sample_c_cd_1d
1466  ; GFX12: bb.1.main_body:
1467  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
1468  ; GFX12-NEXT: {{  $}}
1469  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1470  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1471  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1472  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1473  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1474  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1475  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1476  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1477  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1478  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1479  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1480  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1481  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1482  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1483  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1484  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1485  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1486  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1487  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1488  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1489  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1490  ; GFX12-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1491  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1492  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1493  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1494  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1495  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1496  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
1497  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1498  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1499  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1500  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1501main_body:
1502  %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
1503  ret <4 x float> %v
1504}
1505
1506define amdgpu_ps <4 x float> @sample_c_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t) {
1507  ; GFX10-LABEL: name: sample_c_cd_2d
1508  ; GFX10: bb.1.main_body:
1509  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
1510  ; GFX10-NEXT: {{  $}}
1511  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1512  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1513  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1514  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1515  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1516  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1517  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1518  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1519  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1520  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1521  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1522  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1523  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1524  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1525  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1526  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1527  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1528  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1529  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1530  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1531  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1532  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1533  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
1534  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1535  ; GFX10-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
1536  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1537  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1538  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1539  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1540  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
1541  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1542  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1543  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
1544  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1545  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1546  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1547  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1548  ;
1549  ; GFX11-LABEL: name: sample_c_cd_2d
1550  ; GFX11: bb.1.main_body:
1551  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
1552  ; GFX11-NEXT: {{  $}}
1553  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1554  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1555  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1556  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1557  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1558  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1559  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1560  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1561  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1562  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1563  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1564  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1565  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1566  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1567  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1568  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1569  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1570  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1571  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1572  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1573  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1574  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1575  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
1576  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1577  ; GFX11-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
1578  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1579  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1580  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1581  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1582  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
1583  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1584  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1585  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
1586  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1587  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1588  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1589  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1590  ;
1591  ; GFX12-LABEL: name: sample_c_cd_2d
1592  ; GFX12: bb.1.main_body:
1593  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
1594  ; GFX12-NEXT: {{  $}}
1595  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1596  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1597  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1598  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1599  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1600  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1601  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1602  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1603  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1604  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1605  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1606  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1607  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1608  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1609  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1610  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1611  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1612  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1613  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1614  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1615  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1616  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1617  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
1618  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1619  ; GFX12-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
1620  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1621  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1622  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1623  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1624  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
1625  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
1626  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1627  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1628  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
1629  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1630  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1631  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1632  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1633main_body:
1634  %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
1635  ret <4 x float> %v
1636}
1637
1638define amdgpu_ps <4 x float> @sample_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dsdv, float %s, float %clamp) {
1639  ; GFX10-LABEL: name: sample_cd_cl_1d
1640  ; GFX10: bb.1.main_body:
1641  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
1642  ; GFX10-NEXT: {{  $}}
1643  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1644  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1645  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1646  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1647  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1648  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1649  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1650  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1651  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1652  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1653  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1654  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1655  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1656  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1657  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1658  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1659  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1660  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1661  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1662  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1663  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1664  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1665  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1666  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
1667  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1668  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1669  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1670  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
1671  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1672  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1673  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1674  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1675  ;
1676  ; GFX11-LABEL: name: sample_cd_cl_1d
1677  ; GFX11: bb.1.main_body:
1678  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
1679  ; GFX11-NEXT: {{  $}}
1680  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1681  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1682  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1683  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1684  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1685  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1686  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1687  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1688  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1689  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1690  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1691  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1692  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1693  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1694  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1695  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1696  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1697  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1698  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1699  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1700  ; GFX11-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1701  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1702  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1703  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
1704  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1705  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1706  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1707  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
1708  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1709  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1710  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1711  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1712  ;
1713  ; GFX12-LABEL: name: sample_cd_cl_1d
1714  ; GFX12: bb.1.main_body:
1715  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3
1716  ; GFX12-NEXT: {{  $}}
1717  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1718  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1719  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1720  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1721  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1722  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1723  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1724  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1725  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1726  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1727  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1728  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1729  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1730  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1731  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1732  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1733  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1734  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1735  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1736  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1737  ; GFX12-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1738  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1739  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1740  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY14]](s32)
1741  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1742  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.cl.1d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1743  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1744  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
1745  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1746  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1747  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1748  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1749main_body:
1750  %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.1d.v4f32.f16.f32(i32 15, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
1751  ret <4 x float> %v
1752}
1753
1754define amdgpu_ps <4 x float> @sample_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
1755  ; GFX10-LABEL: name: sample_cd_cl_2d
1756  ; GFX10: bb.1.main_body:
1757  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
1758  ; GFX10-NEXT: {{  $}}
1759  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1760  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1761  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1762  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1763  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1764  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1765  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1766  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1767  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1768  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1769  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1770  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1771  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1772  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1773  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1774  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1775  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1776  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1777  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1778  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1779  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1780  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1781  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1782  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1783  ; GFX10-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
1784  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1785  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1786  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1787  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1788  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
1789  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1790  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1791  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
1792  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1793  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1794  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1795  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1796  ;
1797  ; GFX11-LABEL: name: sample_cd_cl_2d
1798  ; GFX11: bb.1.main_body:
1799  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
1800  ; GFX11-NEXT: {{  $}}
1801  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1802  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1803  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1804  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1805  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1806  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1807  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1808  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1809  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1810  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1811  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1812  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1813  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1814  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1815  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1816  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1817  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1818  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1819  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1820  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1821  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1822  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1823  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1824  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1825  ; GFX11-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
1826  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1827  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1828  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1829  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1830  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
1831  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1832  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1833  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
1834  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1835  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1836  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1837  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1838  ;
1839  ; GFX12-LABEL: name: sample_cd_cl_2d
1840  ; GFX12: bb.1.main_body:
1841  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
1842  ; GFX12-NEXT: {{  $}}
1843  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1844  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1845  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1846  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1847  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1848  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1849  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1850  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1851  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1852  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1853  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1854  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1855  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1856  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1857  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1858  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
1859  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1860  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1861  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1862  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1863  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1864  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
1865  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1866  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
1867  ; GFX12-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
1868  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
1869  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
1870  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1871  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
1872  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
1873  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
1874  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.cd.cl.2d), 15, [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST]](<2 x s16>), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1875  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1876  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
1877  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1878  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1879  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1880  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1881main_body:
1882  %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
1883  ret <4 x float> %v
1884}
1885
1886define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
1887  ; GFX10-LABEL: name: sample_c_cd_cl_1d
1888  ; GFX10: bb.1.main_body:
1889  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
1890  ; GFX10-NEXT: {{  $}}
1891  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1892  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1893  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1894  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1895  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1896  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1897  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1898  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1899  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1900  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1901  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1902  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1903  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1904  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1905  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1906  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1907  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1908  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1909  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1910  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1911  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1912  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1913  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1914  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1915  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1916  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1917  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1918  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.cl.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1919  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1920  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
1921  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1922  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1923  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1924  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1925  ;
1926  ; GFX11-LABEL: name: sample_c_cd_cl_1d
1927  ; GFX11: bb.1.main_body:
1928  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
1929  ; GFX11-NEXT: {{  $}}
1930  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1931  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1932  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1933  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1934  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1935  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1936  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1937  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1938  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1939  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1940  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1941  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1942  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1943  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1944  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1945  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1946  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1947  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1948  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1949  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1950  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1951  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1952  ; GFX11-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1953  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1954  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1955  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1956  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1957  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.cl.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1958  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1959  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
1960  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
1961  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
1962  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
1963  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
1964  ;
1965  ; GFX12-LABEL: name: sample_c_cd_cl_1d
1966  ; GFX12: bb.1.main_body:
1967  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4
1968  ; GFX12-NEXT: {{  $}}
1969  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
1970  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
1971  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
1972  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
1973  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
1974  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
1975  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
1976  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
1977  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
1978  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
1979  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
1980  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
1981  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
1982  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
1983  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
1984  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
1985  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
1986  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
1987  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
1988  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
1989  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
1990  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
1991  ; GFX12-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
1992  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16)
1993  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[DEF]](s16)
1994  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY15]](s32)
1995  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY16]](s32)
1996  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
1997  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.cl.1d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[CONCAT_VECTORS]](<4 x s16>), $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
1998  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
1999  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
2000  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
2001  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
2002  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
2003  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
2004main_body:
2005  %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
2006  ret <4 x float> %v
2007}
2008
2009define amdgpu_ps <4 x float> @sample_c_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {
2010  ; GFX10-LABEL: name: sample_c_cd_cl_2d
2011  ; GFX10: bb.1.main_body:
2012  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
2013  ; GFX10-NEXT: {{  $}}
2014  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
2015  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
2016  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
2017  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
2018  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
2019  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
2020  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
2021  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
2022  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
2023  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
2024  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
2025  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
2026  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
2027  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
2028  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
2029  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
2030  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
2031  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
2032  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
2033  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
2034  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
2035  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
2036  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
2037  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
2038  ; GFX10-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
2039  ; GFX10-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
2040  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
2041  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
2042  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
2043  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
2044  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
2045  ; GFX10-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
2046  ; GFX10-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
2047  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.cl.2d), 15, [[CONCAT_VECTORS]](<12 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
2048  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
2049  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
2050  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
2051  ; GFX10-NEXT:   $vgpr2 = COPY [[UV2]](s32)
2052  ; GFX10-NEXT:   $vgpr3 = COPY [[UV3]](s32)
2053  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
2054  ;
2055  ; GFX11-LABEL: name: sample_c_cd_cl_2d
2056  ; GFX11: bb.1.main_body:
2057  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
2058  ; GFX11-NEXT: {{  $}}
2059  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
2060  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
2061  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
2062  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
2063  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
2064  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
2065  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
2066  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
2067  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
2068  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
2069  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
2070  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
2071  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
2072  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
2073  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
2074  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
2075  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
2076  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
2077  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
2078  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
2079  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
2080  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
2081  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
2082  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
2083  ; GFX11-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
2084  ; GFX11-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
2085  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
2086  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
2087  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
2088  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
2089  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
2090  ; GFX11-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
2091  ; GFX11-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
2092  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.cl.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[CONCAT_VECTORS]](<4 x s16>), $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
2093  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
2094  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
2095  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
2096  ; GFX11-NEXT:   $vgpr2 = COPY [[UV2]](s32)
2097  ; GFX11-NEXT:   $vgpr3 = COPY [[UV3]](s32)
2098  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
2099  ;
2100  ; GFX12-LABEL: name: sample_c_cd_cl_2d
2101  ; GFX12: bb.1.main_body:
2102  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
2103  ; GFX12-NEXT: {{  $}}
2104  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
2105  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
2106  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
2107  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
2108  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
2109  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
2110  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
2111  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
2112  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
2113  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
2114  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
2115  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
2116  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
2117  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
2118  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
2119  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
2120  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY13]](s32)
2121  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
2122  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
2123  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
2124  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
2125  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
2126  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
2127  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
2128  ; GFX12-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
2129  ; GFX12-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
2130  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
2131  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
2132  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
2133  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY17]](s32)
2134  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
2135  ; GFX12-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
2136  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
2137  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.cd.cl.2d), 15, [[BITCAST]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[CONCAT_VECTORS]](<6 x s16>), $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<4 x s32>), addrspace 8)
2138  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
2139  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
2140  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
2141  ; GFX12-NEXT:   $vgpr2 = COPY [[UV2]](s32)
2142  ; GFX12-NEXT:   $vgpr3 = COPY [[UV3]](s32)
2143  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3
2144main_body:
2145  %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
2146  ret <4 x float> %v
2147}
2148
2149define amdgpu_ps float @sample_c_d_o_2darray_V1(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice) {
2150  ; GFX10-LABEL: name: sample_c_d_o_2darray_V1
2151  ; GFX10: bb.1.main_body:
2152  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
2153  ; GFX10-NEXT: {{  $}}
2154  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
2155  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
2156  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
2157  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
2158  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
2159  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
2160  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
2161  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
2162  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
2163  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
2164  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
2165  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
2166  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
2167  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
2168  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
2169  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
2170  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
2171  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
2172  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
2173  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
2174  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
2175  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
2176  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
2177  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
2178  ; GFX10-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
2179  ; GFX10-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
2180  ; GFX10-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
2181  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
2182  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
2183  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
2184  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
2185  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
2186  ; GFX10-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
2187  ; GFX10-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY20]](s32)
2188  ; GFX10-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<14 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
2189  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.o.2darray), 4, [[CONCAT_VECTORS]](<14 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (s32), addrspace 8)
2190  ; GFX10-NEXT:   $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
2191  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0
2192  ;
2193  ; GFX11-LABEL: name: sample_c_d_o_2darray_V1
2194  ; GFX11: bb.1.main_body:
2195  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
2196  ; GFX11-NEXT: {{  $}}
2197  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
2198  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
2199  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
2200  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
2201  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
2202  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
2203  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
2204  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
2205  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
2206  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
2207  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
2208  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
2209  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
2210  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
2211  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
2212  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
2213  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
2214  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
2215  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
2216  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
2217  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
2218  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
2219  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
2220  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
2221  ; GFX11-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
2222  ; GFX11-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
2223  ; GFX11-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
2224  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
2225  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
2226  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
2227  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
2228  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
2229  ; GFX11-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
2230  ; GFX11-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY20]](s32)
2231  ; GFX11-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
2232  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.o.2darray), 4, [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[CONCAT_VECTORS]](<6 x s16>), $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (s32), addrspace 8)
2233  ; GFX11-NEXT:   $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
2234  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0
2235  ;
2236  ; GFX12-LABEL: name: sample_c_d_o_2darray_V1
2237  ; GFX12: bb.1.main_body:
2238  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
2239  ; GFX12-NEXT: {{  $}}
2240  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
2241  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
2242  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
2243  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
2244  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
2245  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
2246  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
2247  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
2248  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
2249  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
2250  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
2251  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
2252  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
2253  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
2254  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
2255  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
2256  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
2257  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
2258  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
2259  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
2260  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
2261  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
2262  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
2263  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
2264  ; GFX12-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
2265  ; GFX12-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
2266  ; GFX12-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
2267  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
2268  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
2269  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
2270  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
2271  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
2272  ; GFX12-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
2273  ; GFX12-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY20]](s32)
2274  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
2275  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.o.2darray), 4, [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[CONCAT_VECTORS]](<8 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (s32), addrspace 8)
2276  ; GFX12-NEXT:   $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](s32)
2277  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0
2278main_body:
2279  %v = call float @llvm.amdgcn.image.sample.c.d.o.2darray.f16.f32.f32(i32 4, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
2280  ret float %v
2281}
2282
2283define amdgpu_ps <2 x float> @sample_c_d_o_2darray_V2(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice) {
2284  ; GFX10-LABEL: name: sample_c_d_o_2darray_V2
2285  ; GFX10: bb.1.main_body:
2286  ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
2287  ; GFX10-NEXT: {{  $}}
2288  ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
2289  ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
2290  ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
2291  ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
2292  ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
2293  ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
2294  ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
2295  ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
2296  ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
2297  ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
2298  ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
2299  ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
2300  ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
2301  ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
2302  ; GFX10-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
2303  ; GFX10-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
2304  ; GFX10-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
2305  ; GFX10-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
2306  ; GFX10-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
2307  ; GFX10-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
2308  ; GFX10-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
2309  ; GFX10-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
2310  ; GFX10-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
2311  ; GFX10-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
2312  ; GFX10-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
2313  ; GFX10-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
2314  ; GFX10-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
2315  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
2316  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
2317  ; GFX10-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
2318  ; GFX10-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
2319  ; GFX10-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
2320  ; GFX10-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
2321  ; GFX10-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY20]](s32)
2322  ; GFX10-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<14 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
2323  ; GFX10-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.o.2darray), 6, [[CONCAT_VECTORS]](<14 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<2 x s32>), addrspace 8)
2324  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
2325  ; GFX10-NEXT:   $vgpr0 = COPY [[UV]](s32)
2326  ; GFX10-NEXT:   $vgpr1 = COPY [[UV1]](s32)
2327  ; GFX10-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
2328  ;
2329  ; GFX11-LABEL: name: sample_c_d_o_2darray_V2
2330  ; GFX11: bb.1.main_body:
2331  ; GFX11-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
2332  ; GFX11-NEXT: {{  $}}
2333  ; GFX11-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
2334  ; GFX11-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
2335  ; GFX11-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
2336  ; GFX11-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
2337  ; GFX11-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
2338  ; GFX11-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
2339  ; GFX11-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
2340  ; GFX11-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
2341  ; GFX11-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
2342  ; GFX11-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
2343  ; GFX11-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
2344  ; GFX11-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
2345  ; GFX11-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
2346  ; GFX11-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
2347  ; GFX11-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
2348  ; GFX11-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
2349  ; GFX11-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
2350  ; GFX11-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
2351  ; GFX11-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
2352  ; GFX11-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
2353  ; GFX11-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
2354  ; GFX11-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
2355  ; GFX11-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
2356  ; GFX11-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
2357  ; GFX11-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
2358  ; GFX11-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
2359  ; GFX11-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
2360  ; GFX11-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
2361  ; GFX11-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
2362  ; GFX11-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
2363  ; GFX11-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
2364  ; GFX11-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
2365  ; GFX11-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
2366  ; GFX11-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY20]](s32)
2367  ; GFX11-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
2368  ; GFX11-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.o.2darray), 6, [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[CONCAT_VECTORS]](<6 x s16>), $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<2 x s32>), addrspace 8)
2369  ; GFX11-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
2370  ; GFX11-NEXT:   $vgpr0 = COPY [[UV]](s32)
2371  ; GFX11-NEXT:   $vgpr1 = COPY [[UV1]](s32)
2372  ; GFX11-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
2373  ;
2374  ; GFX12-LABEL: name: sample_c_d_o_2darray_V2
2375  ; GFX12: bb.1.main_body:
2376  ; GFX12-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
2377  ; GFX12-NEXT: {{  $}}
2378  ; GFX12-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
2379  ; GFX12-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
2380  ; GFX12-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
2381  ; GFX12-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
2382  ; GFX12-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
2383  ; GFX12-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
2384  ; GFX12-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
2385  ; GFX12-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
2386  ; GFX12-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
2387  ; GFX12-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $sgpr10
2388  ; GFX12-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $sgpr11
2389  ; GFX12-NEXT:   [[COPY10:%[0-9]+]]:_(s32) = COPY $sgpr12
2390  ; GFX12-NEXT:   [[COPY11:%[0-9]+]]:_(s32) = COPY $sgpr13
2391  ; GFX12-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
2392  ; GFX12-NEXT:   [[COPY12:%[0-9]+]]:_(s32) = COPY $vgpr0
2393  ; GFX12-NEXT:   [[COPY13:%[0-9]+]]:_(s32) = COPY $vgpr1
2394  ; GFX12-NEXT:   [[COPY14:%[0-9]+]]:_(s32) = COPY $vgpr2
2395  ; GFX12-NEXT:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY14]](s32)
2396  ; GFX12-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY $vgpr3
2397  ; GFX12-NEXT:   [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY15]](s32)
2398  ; GFX12-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY $vgpr4
2399  ; GFX12-NEXT:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY16]](s32)
2400  ; GFX12-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY $vgpr5
2401  ; GFX12-NEXT:   [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY17]](s32)
2402  ; GFX12-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY $vgpr6
2403  ; GFX12-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr7
2404  ; GFX12-NEXT:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr8
2405  ; GFX12-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY12]](s32)
2406  ; GFX12-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY13]](s32)
2407  ; GFX12-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
2408  ; GFX12-NEXT:   [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
2409  ; GFX12-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY18]](s32)
2410  ; GFX12-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY19]](s32)
2411  ; GFX12-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[COPY20]](s32)
2412  ; GFX12-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR3]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
2413  ; GFX12-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.c.d.o.2darray), 6, [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[CONCAT_VECTORS]](<8 x s16>), $noreg, $noreg, $noreg, $noreg, $noreg, [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 2 :: (dereferenceable load (<2 x s32>), addrspace 8)
2414  ; GFX12-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
2415  ; GFX12-NEXT:   $vgpr0 = COPY [[UV]](s32)
2416  ; GFX12-NEXT:   $vgpr1 = COPY [[UV1]](s32)
2417  ; GFX12-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
2418main_body:
2419  %v = call <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f16.f32(i32 6, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
2420  ret <2 x float> %v
2421}
2422
2423declare <4 x float> @llvm.amdgcn.image.sample.d.1d.v4f32.f16.f32(i32, half, half, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2424declare <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f32(i32, half, half, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2425declare <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f16.f32(i32, half, half, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2426declare <4 x float> @llvm.amdgcn.image.sample.c.d.1d.v4f32.f16.f32(i32, float, half, half, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2427declare <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f16.f32(i32, float, half, half, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2428declare <4 x float> @llvm.amdgcn.image.sample.d.cl.1d.v4f32.f16.f32(i32, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2429declare <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2430declare <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f16.f32(i32, float, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2431declare <4 x float> @llvm.amdgcn.image.sample.c.d.cl.2d.v4f32.f16.f32(i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2432
2433declare <4 x float> @llvm.amdgcn.image.sample.cd.1d.v4f32.f16.f32(i32, half, half, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2434declare <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f32(i32, half, half, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2435declare <4 x float> @llvm.amdgcn.image.sample.c.cd.1d.v4f32.f16.f32(i32, float, half, half, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2436declare <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f16.f32(i32, float, half, half, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2437declare <4 x float> @llvm.amdgcn.image.sample.cd.cl.1d.v4f32.f16.f32(i32, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2438declare <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2439declare <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32, float, half, half, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2440declare <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f16.f32(i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2441
2442declare float @llvm.amdgcn.image.sample.c.d.o.2darray.f16.f32.f32(i32, i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2443declare <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f16.f32(i32, i32, float, half, half, half, half, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
2444
2445attributes #0 = { nounwind }
2446attributes #1 = { nounwind readonly }
2447attributes #2 = { nounwind readnone }
2448